Low rate hippocampal delay period activity encodes behavioral experience

Markos Athanasiadis, Stefano Masserini, Li Yuan, Dustin Fetterhoff, Jill K. Leutgeb, Stefan Leutgeb, Christian Leibold

TL;DR

This paper reveals that the hippocampus encodes recent experiences using surprisingly sparse and low-rate neural activity, occurring in brief, infrequent bursts. This "whisper" of information, rather than a sustained "shout," is crucial for working memory and is present even when explicit memory isn't required, suggesting a fundamental role in information processing.

What Problem Does This Paper Address?

The precise neural mechanisms underlying short-term memory, particularly how the brain retains information about recent events for brief periods, remain elusive. While persistent neural firing has been a leading hypothesis, experimental evidence has been mixed, especially in the hippocampus. This study aims to uncover how the hippocampus encodes and maintains information about recent sensory-motor experiences, even when explicit working memory is not demanded, and to identify the specific neural activity patterns responsible.

Methods

The researchers employed a sophisticated classifier-based decoding approach. They trained a linear neural network to predict past behavioral choices (left/right turns) from time-binned hippocampal neural activity. To understand which neurons and when this information was encoded, they utilized "adversarial attacks" – a machine learning technique to probe the decision boundary of the classifier. This allowed them to identify "Most Informative Directions" (MIDs), representing neural ensembles crucial for decoding. They then analyzed "relevant time bins" – specific moments where neural activity strongly correlated with MIDs – to characterize the nature of the informative signals.

Datasets

The study analyzed two previously published electrophysiological datasets:

  1. Rat data: Recordings from CA1 and CA3 regions of rats performing a spatial alternation task with a delayed choice. This dataset included control rats and rats with lesions to the medial entorhinal cortex (mEC).
  2. Gerbil data: Recordings from CA1 of Mongolian gerbils navigating virtual reality mazes. This dataset included both familiar and "swapped" mazes (where visual cues were associated with different turns, introducing sensory conflict), and the animals were not required to perform an explicit working memory task.

Key Insights and Ideas

  • The "Whisper" of Information: The most striking finding is that hippocampal information about recent experiences is encoded not by sustained, high-rate firing, but by brief, infrequent bursts of activity involving only a small number of neurons. This suggests a highly energy-efficient mechanism for short-term memory.
  • Beyond Explicit Working Memory: The informative low-rate patterns were observed even in tasks that did not explicitly require working memory, implying this mechanism is a fundamental aspect of hippocampal information processing, not solely tied to demanding recall tasks.
  • mEC's Role: Lesions to the medial entorhinal cortex significantly reduced the expression of these informative patterns, particularly in CA3, highlighting the mEC's crucial role in supporting this hippocampal encoding mechanism.
  • Sensory Context Matters: The ability to decode past experiences was diminished in gerbils navigating mazes with conflicting sensory cues, suggesting that the familiarity and consistency of the sensory-motor context influence the clarity of these hippocampal traces.

Key Results

  • A linear neural network could decode previous trial outcomes from hippocampal activity with statistically significant accuracy in a fraction of recorded sessions.
  • This decoding was most effective in short time bins (100 ms), indicating that the informative signals are transient.
  • The informative activity was sparse, with only a small percentage of neurons (around 20%) and time bins (around 2%) contributing significantly.
  • The firing rates in these "relevant" time bins were only slightly elevated compared to non-relevant bins.
  • mEC lesions reduced the number of relevant time bins and the overall decoding accuracy, especially in CA3.
  • Gerbils in "swapped" mazes showed significantly fewer relevant time bins and lower decoding accuracy compared to familiar mazes.

Limitations and Future Work

The authors acknowledge that the statistical power of their multivariate classifier approach is limited by the number of trials and recording session length. They also note that their method is tuned to detect information in the covariance structure of activity, potentially missing smaller, but still significant, rate effects observed in previous studies.

Future work could focus on understanding how this low-rate information is maintained over longer periods (tens of seconds to minutes) and how it is transferred to downstream brain regions. Investigating the precise synaptic mechanisms underlying these brief, informative bursts and exploring their role in other cognitive functions beyond spatial memory are also promising avenues. The authors suggest that further analysis approaches are needed to identify these neural signatures, which are challenging to find due to their sparse and low-rate nature.

Summary updated

Abstract

Remembering what just happened is a crucial prerequisite to form long-term memories but also for establishing and maintaining working memory. So far there is no general agreement about cortical mechanisms that support short-term memory. Using a classifier-based decoding approach, we report that hippocampal activity during few sparsely distributed brief time intervals contains information about the previous sensory motor experience of rodents. These intervals are characterized by only a small increase of firing rate of only a few neurons. These low-rate predictive patterns are present in both working memory and non-working memory tasks, in two rodent species, rats and Mongolian gerbils, are strongly reduced for rats with medial entorhinal cortex lesions, and depend on the familiarity of the sensory-motor context.
BibTeX
@inproceedings{athanasiadis2024low,
  author = {Markos Athanasiadis and Stefano Masserini and Li Yuan and Dustin Fetterhoff and Jill K. Leutgeb and Stefan Leutgeb and Christian Leibold},
  title = {{Low rate hippocampal delay period activity encodes behavioral experience}},
  booktitle = {Hippocampus},
  year = {2024},
  doi = {10.1002/hipo.23619},
}
Figure